Information for prospective students
Research topics available
We are currently looking for Master's and Bachelor students, who would be interested in research topics on coding and image compression.
- Context-based recompression of JPEG coded images
The JPEG (Joint Experts Pictures Group) standard was adopted in 1992. In spite of existence of more efficient standards for image compression (for example, JPEG2000 and JPEG-XR), nowadays JPEG is one of the most popular digital standards used for representing colour images in a compact form. One of the reasons for its popularity is that JPEG compression is commonly used in digital photo cameras. Practically every modern digital camera has inbuilt compression algorithm which stores images in JPEG. With improving resolutions of the digital photos, it becomes more and more difficult to find on the market a digital camera using so-called raw format, that is, without compression.
There exists a necessity to store a large number of existing digital photos, which generates a problem of representing them in a more compact form. Since JPEG compression causes information loss, a straightforward approach based on reconstruction of the original image followed by compression using more efficient standards is not applicable. On the other hand, universal lossless coding techniques which do not take into account statistical properties of images are also not efficient. Context-based arithmetic coding is one of the most efficient lossless compression techniques which can be efficiently used for recompression JPEG coded images.
In this project, the student will develop and implement an efficient algorithm for recompression of JPEG coded images using context-based arithmetic coding.
- Packet loss correction in vehicle communication networks
The development of vehicle communication networks, which will exclude (mostly and in the future completely) a person from the driving process, is one of the priority areas in the field of communication. Providing traffic safety in vehicle-to-vehicle communication networks imposes high reliability, low delay and complexity requirements on packets exchange.
In the existing standards for V2V networks, information is transmitted in packets (MACs) of length about a few thousands of bits. In particular, geographical location data, direction of movement as well as some emergency information about fast changes in the traffic situation are transmitted. The goal of the project is to develop coding and decoding procedures for the packet level of networks taking into account the above mentioned restrictions. One more important fact which should be taken into account is that a significant part of the transmitted data plays an extremely important role in ensuring road safety but in a short time becomes completely
useless.
- Video and image compression for networks
The development of vehicle communication networks, which will exclude (mostly and in the future completely) a person from the driving process, is one of the priority areas in the field of communication. In such networks vehicles exchange video data but errors in the communication channel can lead to the loss of video data packets. Reliability of communication can be improved by using error correcting codes but at the cost of transmission rate reduction.
The goal of the project is to develop compression-coding system providing the best reliability-transmission rate tradeoff.
- Error correcting codes for 5G
Optimization of code parameters is in demand by technology companies competing for leadership in developing new generation communication standards. Low density parity-check (LDPC) codes are among code candidates to the future standards. An attractive feature of these codes is existence of low complexity iterative decoding procedure applicable to them.
The goal of the project is to choose code structure taking into account limitations on the implementation complexity of decoding algorithms.
- Wavelet based image denoising
Digital images can often be contaminated by noise. There are different reasons for this phenomenon, for example, an image can be corrupted by noise during transmission over a communication channel or a source of image can produce a mixture of image and noise due to specific features of image generating devices etc. For a number of applications it is extremely important to remove noise preserving useful information in the image (for example, medical images). Wavelet filtering is a decomposition of the original image into high- and low-frequency components. Then denoising techniques are applied to the obtained image components.
The goal of the project is to remove noise and improve image quality.
- Context-based lossless image compression
A specific feature of any digital image is a rather high correlation between its pixels. This circumstance leads to inefficiency of any independent processing of the pixels. This problem can be partially solved by applying to the image various prediction techniques followed by context coding, that is, splitting image data into statistically similar areas. Then to these areas lossless coding techniques such as, for example, arithmetic coding are applied.
The goal of the project is improving efficiency of lossless image coding by searching for the new system of contexts (rules for data splitting).
Requirements
-
good academic records in computer science, electronics engineering, physics or related discipline
-
good mathematical, analytical and programming skills
All projects can be first taken as MTAT.07.027 Special Assignment in Coding Theory and later extended to Master's or Bachelor thesis.
Interested students should get in touch with us:
irina.bocharova at ut.ee, boris.kudryashov at ut.ee, vitaly.skachek at ut.ee